Re: [RFC][PATCH] inotify 0.8.1 [u]



On Tue, 2004-08-24 at 03:04, John McCutchan wrote:

> > Which then results in a panic as the dm volumes cannot be setup
> > and no / found by kernel.  So basically it seems like inotify
> > mess with dm in some way or other - any quick ideas what it
> > could be?
> > 
> 
> This is very strange. 'inotify device opened' is printed when an app
> opens /dev/inotify. Do you have code that is doing that? I am wondering
> if inotify's MAJOR/MINOR is getting confused with device mappers. I
> don't know how it could be happening since /dev/inotify has the kernel
> assign it an available minor number, and its major number is the major
> number for all misc char devices.
> 

Ok, seems like this was my own stupidity - I used an static node in
initramfs which the kernel assigned the minor of to inotify if compiled
in (I did not know DM used dynamic minors) :/  Fixed it to use udev
instead.

I am wondering about two things though:
 1) Should things like device-mapper that are many times boot critical
use dynamic minors?  Guess if we make an exception for one, somebody
will always get some silly device to be boot critical :(  So I'll
imagine it should stay as is - maybe add some nice fat message
somewhere?
 2) Should ash from klibc test the file (if [ -s /foo/bar]) a symlink
points to instead of the symlink itself?

---
nosferatu klibc # ls -l /dev/isw0
lrwxrwxrwx  1 root root 11 Aug 25  2004 /dev/isw0 -> mapper/dm-0
nosferatu klibc # ./ash/sh -c 'if [ -s /dev/isw0 ]; then echo yes; fi'
nosferatu klibc # ./ash/sh -c 'if [ -b /dev/isw0 ]; then echo yes; fi'
yes
nosferatu klibc #
---

As a note - I had to do a 'while [ ! -b /dev/isw0 ]; do sleep 1; done'
as 'node creation latency' with udev is an second or two even on a
p4 3ghz.


Thanks,

-- 
Martin Schlemmer

Attachment: signature.asc
Description: This is a digitally signed message part



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]